How Google Ranks Pages

The Ranking Pipeline: Not One Algorithm, But Many

Most SEO practitioners picture Google ranking as a single formula that evaluates a page and assigns it a position. The reality is far more complex. A URL first receives a base relevance score, then passes through a series of re-ranking systems called Twiddlers — internal mechanisms that apply user-quality signals on top of raw relevance.

This pipeline architecture explains a counterintuitive observation: a page can be well-optimized for a keyword and initially appear in the top 10, then gradually lose visibility over weeks. The initial ranking reflects relevance signals. The subsequent drop reflects quality signals that only emerge once real users interact with the page.

Site Authority: The Domain-Level Trust Signal

Google computes two critical site-level authority metrics that influence every page on a domain:

• siteAuthority — a site-level score that provides the trust context for all pages on the domain. This is Google's internal equivalent of what the SEO industry calls "domain authority."
• NSR (Normalized Site Rank) — an advanced quality and reliability metric computed from multiple signals. It's one of the primary inputs to the overall domain quality score.

These metrics mean that a new, high-quality page on an established domain will outperform the same page on a new domain, even with identical on-page optimization. Building siteAuthority requires time, consistent quality content, and legitimate backlinks — there are no shortcuts.

Behavioral Signals: GoodClicks, BadClicks, LastLongestClicks

Google records detailed data about how users interact with search results. Three signals are particularly important:

• GoodClicks — users click your result and do not return to the SERP (they found what they needed)
• BadClicks — users click and immediately return to search (pogo-sticking), signaling the page failed to satisfy the query
• LastLongestClicks — the final click in a search session, indicating the user's search was fully resolved on that page

These are not theoretical signals — they are direct inputs to re-ranking systems. A page with consistently high BadClicks will gradually lose positions regardless of how well-optimized its HTML is. Conversely, a page that consistently earns LastLongestClicks receives a strong quality signal and tends to hold or improve positions.

Freshness: More Than a Publication Date

Google's freshness evaluation goes far beyond checking when a page was published. The system uses a signal called lastSignificantUpdate to distinguish genuine content improvements from cosmetic edits (changing a date in the footer or adding a single sentence). Only substantial content changes trigger freshness signals.

The system rewards real improvements: adding new data, covering recent events, restructuring outdated sections. Simply changing the publication date without meaningful content changes provides no freshness benefit and may be flagged as manipulation.

From Keywords to Entities and Embeddings

Google's ranking architecture has fundamentally shifted from keyword matching to semantic understanding. Two key systems underpin this:

• EntityAnnotations — the system identifies entities in both queries and documents, matches them, and scores relevance based on entity coverage and relationships
• site2vecEmbeddingEncoded — machine-learning vector embeddings computed for the entire site, representing its semantic "meaning" as a vector in high-dimensional space

This means topical authority is now mathematically computable. Google can calculate how close a site's semantic vector is to the query's vector, effectively measuring "how much of an expert is this domain on this topic." Keyword density is irrelevant; entity coverage and depth of topic treatment are what matter.

Internal Link Prominence: onsiteProminence

The metric onsiteProminence measures a page's importance within the site by simulating user traffic flow from the homepage and other high-traffic pages. This is essentially internal PageRank — the more internal links point to a page from high-authority pages, the more prominent it is in Google's internal model.

The practical implication: your internal linking strategy directly determines which pages Google considers most important on your site. Pages buried without internal links, even with strong content, receive low onsiteProminence scores and rank harder.